Parametrization of Quantificational Determiners and Head-Internal Relatives

This paper proposes that quantificational determiners come in two varieties, one which makes use of feature checking and the other that relies on binding, and that the typology of head-internal relative clauses (HIRC) is sensitive to this parametrization of quantificational determiners, once the existence of HIRC is made possible by an independent parameter. Challenge: Our point of departure is the observation by Kondo 81 that HIRC, absent in the Nara Period, appeared in Old Japanese during the Heian Period. This change correlates with the loss of wh-movement (Watanabe to appear) during the Heian Period, which supports the line of analysis that invokes essentially the same movement/feature-checking mechanism for HIRC as for wh-in-situ, as in Watanabe 92. This idea, however, is challenged by languages like Chinese on the one hand, and by languages like Imbabura Quechua, on the other hand. Chinese has wh-in-situ but lacks HIRC. Imbabura Quechua has obligatory wh-movement but allows HIRC. What is going on? Aoun & Li (1993) argue that the Chinese problem is solved by observing that Chinese lacks quantificational particles that attach to wh-indefinites in languages like Modern Japanese. Assuming that a similar, albeit abstract particle is responsible for movement/feature-checking of wh-in-situ and HIRC, the absence of HIRC in Chinese follows. But this cannot be the end of the story, because there is a variety of HIRC which does not involve movement or feature-checking but relies on unselective binding. Lakhota HIRC, which does not exhibit island effects (Williamson 87), is analyzed by Bonneau 92 as making use of unselective binding. Why can't Chinese choose the unselective binding strategy to make HIRC possible? Bonneau points to the determiner that comes with HIRC (as well as with ordinary NP) in Lakhota as the unselective binder. The account of the absence of HIRC in Chinese now seems complete, since Chinese lacks the determiner system. At this point, however, we should ask a deeper question of why things are as they are. That is, why must quantificational particles induce movement/feature-checking while determiners are unselective binders? If things were the other way round, we would find Lakhota HIRC sensitive to islands and Modern Japanese HIRC violating subjacency, contrary to fact. We need a more principled answer. Typology of determiners: I would like to propose that quantificational particles are determiners after all, and that binding and feature checking are the two modes provided by UG to form association of quantificational determiners with nominal predicates which is needed for quantification. So, the contrast between Modern Japanese and Lakhota with respect to HIRC is now reformulated as the question of parametrization of determiners. Treating quantificational particles as determiners points us to an interesting difference between the two types of quantificational determiners: quantificational particles in Modern Japanese are selective whereas determiners in Lakhota are not. Thus, the particle ka in Modern Japanese, which forms an existential quantifier, can only attach to wh-indefinites, as in (1). (1) a. Dare-ka-ga kita. "Someone came." who-KA-nom came b. *Otoko-ka-ga kita. "Some man came." man-KA-nom (1b) is ill-formed, because the particle ka cannot combine with an ordinary NP. Determiners in Lakhota, on the other hand, do not show such a restriction. The examples in (2) illustrate the